![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to tcp_input.c CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [Previous NeXT emulator] / previous_trunk / src / slirp|
Return to tcp_input.c CVS log | Up to [Previous NeXT emulator] / previous_trunk / src / slirp |
1.1 root 1: /*
2: * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994
3: * The Regents of the University of California. All rights reserved.
4: *
5: * Redistribution and use in source and binary forms, with or without
6: * modification, are permitted provided that the following conditions
7: * are met:
8: * 1. Redistributions of source code must retain the above copyright
9: * notice, this list of conditions and the following disclaimer.
10: * 2. Redistributions in binary form must reproduce the above copyright
11: * notice, this list of conditions and the following disclaimer in the
12: * documentation and/or other materials provided with the distribution.
13: * 3. Neither the name of the University nor the names of its contributors
14: * may be used to endorse or promote products derived from this software
15: * without specific prior written permission.
16: *
17: * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20: * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27: * SUCH DAMAGE.
28: *
29: * @(#)tcp_input.c 8.5 (Berkeley) 4/10/94
30: * tcp_input.c,v 1.10 1994/10/13 18:36:32 wollman Exp
31: */
32:
33: /*
34: * Changes and additions relating to SLiRP
35: * Copyright (c) 1995 Danny Gasparovski.
36: *
37: * Please read the file COPYRIGHT for the
38: * terms and conditions of the copyright.
39: */
40:
41: #include <stdlib.h>
42: #include <slirp.h>
43: #include "ip_icmp.h"
44:
45: struct socket tcb;
46:
47: int tcprexmtthresh = 3;
48: struct socket *tcp_last_so = &tcb;
49:
50: tcp_seq tcp_iss; /* tcp initial send seq # */
51:
52: #define TCP_PAWS_IDLE (24 * 24 * 60 * 60 * PR_SLOWHZ)
53:
54: /* for modulo comparisons of timestamps */
55: #define TSTMP_LT(a,b) ((int)((a)-(b)) < 0)
56: #define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0)
57:
58: /*
59: * Insert segment ti into reassembly queue of tcp with
60: * control block tp. Return TH_FIN if reassembly now includes
61: * a segment with FIN. The macro form does the common case inline
62: * (segment is the next to be received on an established connection,
63: * and the queue is empty), avoiding linkage into and removal
64: * from the queue and repetition of various conversions.
65: * Set DELACK for segments received in order, but ack immediately
66: * when segments are out of order (so fast retransmit can work).
67: */
68: #ifdef TCP_ACK_HACK
69: #define TCP_REASS(tp, ti, m, so, flags) {\
70: if ((ti)->ti_seq == (tp)->rcv_nxt && \
71: tcpfrag_list_empty(tp) && \
72: (tp)->t_state == TCPS_ESTABLISHED) {\
73: if (ti->ti_flags & TH_PUSH) \
74: tp->t_flags |= TF_ACKNOW; \
75: else \
76: tp->t_flags |= TF_DELACK; \
77: (tp)->rcv_nxt += (ti)->ti_len; \
78: flags = (ti)->ti_flags & TH_FIN; \
79: tcpstat.tcps_rcvpack++;\
80: tcpstat.tcps_rcvbyte += (ti)->ti_len;\
81: if (so->so_emu) { \
82: if (tcp_emu((so),(m))) sbappend((so), (m)); \
83: } else \
84: sbappend((so), (m)); \
85: /* sorwakeup(so); */ \
86: } else {\
87: (flags) = tcp_reass((tp), (ti), (m)); \
88: tp->t_flags |= TF_ACKNOW; \
89: } \
90: }
91: #else
92: #define TCP_REASS(tp, ti, m, so, flags) { \
93: if ((ti)->ti_seq == (tp)->rcv_nxt && \
94: tcpfrag_list_empty(tp) && \
95: (tp)->t_state == TCPS_ESTABLISHED) { \
96: tp->t_flags |= TF_DELACK; \
97: (tp)->rcv_nxt += (ti)->ti_len; \
98: flags = (ti)->ti_flags & TH_FIN; \
99: tcpstat.tcps_rcvpack++;\
100: tcpstat.tcps_rcvbyte += (ti)->ti_len;\
101: if (so->so_emu) { \
102: if (tcp_emu((so),(m))) sbappend(so, (m)); \
103: } else \
104: sbappend((so), (m)); \
105: /* sorwakeup(so); */ \
106: } else { \
107: (flags) = tcp_reass((tp), (ti), (m)); \
108: tp->t_flags |= TF_ACKNOW; \
109: } \
110: }
111: #endif
112:
113: int
114: tcp_reass(register struct tcpcb *tp, register struct tcpiphdr *ti, struct mbuf *m)
115: {
116: register struct tcpiphdr *q;
117: struct socket *so = tp->t_socket;
118: int flags;
119:
120: /*
121: * Call with ti==0 after become established to
122: * force pre-ESTABLISHED data up to user socket.
123: */
124: if (ti == 0)
125: goto present;
126:
127: /*
128: * Find a segment which begins after this one does.
129: */
130: for (q = tcpfrag_list_first(tp); !tcpfrag_list_end(q, tp);
131: q = tcpiphdr_next(q))
132: if (SEQ_GT(q->ti_seq, ti->ti_seq))
133: break;
134:
135: /*
136: * If there is a preceding segment, it may provide some of
137: * our data already. If so, drop the data from the incoming
138: * segment. If it provides all of our data, drop us.
139: */
140: if (!tcpfrag_list_end(tcpiphdr_prev(q), tp)) {
141: register int i;
142: q = tcpiphdr_prev(q);
143: /* conversion to int (in i) handles seq wraparound */
144: i = q->ti_seq + q->ti_len - ti->ti_seq;
145: if (i > 0) {
146: if (i >= ti->ti_len) {
147: tcpstat.tcps_rcvduppack++;
148: tcpstat.tcps_rcvdupbyte += ti->ti_len;
149: m_freem(m);
150: /*
151: * Try to present any queued data
152: * at the left window edge to the user.
153: * This is needed after the 3-WHS
154: * completes.
155: */
156: goto present; /* ??? */
157: }
158: m_adj(m, i);
159: ti->ti_len -= i;
160: ti->ti_seq += i;
161: }
162: q = tcpiphdr_next(q);
163: }
164: tcpstat.tcps_rcvoopack++;
165: tcpstat.tcps_rcvoobyte += ti->ti_len;
166: ti->ti_mbuf = m;
167:
168: /*
169: * While we overlap succeeding segments trim them or,
170: * if they are completely covered, dequeue them.
171: */
172: while (!tcpfrag_list_end(q, tp)) {
173: register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
174: if (i <= 0)
175: break;
176: if (i < q->ti_len) {
177: q->ti_seq += i;
178: q->ti_len -= i;
179: m_adj(q->ti_mbuf, i);
180: break;
181: }
182: q = tcpiphdr_next(q);
183: m = tcpiphdr_prev(q)->ti_mbuf;
184: remque(tcpiphdr2qlink(tcpiphdr_prev(q)));
185: m_freem(m);
186: }
187:
188: /*
189: * Stick new segment in its place.
190: */
191: insque(tcpiphdr2qlink(ti), tcpiphdr2qlink(tcpiphdr_prev(q)));
192: present:
193: /*
194: * Present data to user, advancing rcv_nxt through
195: * completed sequence space.
196: */
197: if (!TCPS_HAVEESTABLISHED(tp->t_state))
198: return (0);
199: ti = tcpfrag_list_first(tp);
200: if (tcpfrag_list_end(ti, tp) || ti->ti_seq != tp->rcv_nxt)
201: return (0);
202: if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
203: return (0);
204: do {
205: tp->rcv_nxt += ti->ti_len;
206: flags = ti->ti_flags & TH_FIN;
207: remque(tcpiphdr2qlink(ti));
208: m = ti->ti_mbuf;
209: ti = tcpiphdr_next(ti);
210: /* if (so->so_state & SS_FCANTRCVMORE) */
211: if (so->so_state & SS_FCANTSENDMORE)
212: m_freem(m);
213: else {
214: if (so->so_emu) {
215: if (tcp_emu(so,m)) sbappend(so, m);
216: } else
217: sbappend(so, m);
218: }
219: } while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
220: /* sorwakeup(so); */
221: return (flags);
222: }
223:
224: /*
225: * TCP input routine, follows pages 65-76 of the
226: * protocol specification dated September, 1981 very closely.
227: */
228: void tcp_input(register struct mbuf *m, int iphlen, struct socket *inso)
229: {
230: struct ip save_ip, *ip;
231: register struct tcpiphdr *ti;
232: caddr_t optp = NULL;
233: int optlen = 0;
234: int len, tlen, off;
235: register struct tcpcb *tp = 0;
236: register int tiflags;
237: struct socket *so = 0;
238: int todrop;
239: u_int acked;
240: int ourfinisacked, needoutput = 0;
241: /* int dropsocket = 0; */
242: int iss = 0;
243: u_long tiwin;
244: int ret;
245: /* int ts_present = 0; */
246:
247: DEBUG_CALL("tcp_input");
248: DEBUG_ARGS((dfd, " m = %8lx iphlen = %2d inso = %lx\n",
249: (long)m, iphlen, (long)inso));
250:
251: /*
252: * If called with m == 0, then we're continuing the connect
253: */
254: if (m == NULL) {
255: so = inso;
256:
257: /* Re-set a few variables */
258: tp = sototcpcb(so);
259: m = so->so_m;
260: so->so_m = 0;
261: ti = so->so_ti;
262: tiwin = ti->ti_win;
263: tiflags = ti->ti_flags;
264:
265: goto cont_conn;
266: }
267:
268:
269: tcpstat.tcps_rcvtotal++;
270: /*
271: * Get IP and TCP header together in first mbuf.
272: * Note: IP leaves IP header in first mbuf.
273: */
274: ti = mtod(m, struct tcpiphdr *);
275: if (iphlen > sizeof(struct ip)) {
276: ip_stripoptions(m, (struct mbuf *)0);
277: iphlen = sizeof(struct ip);
278: }
279: /* XXX Check if too short */
280:
281:
282: /*
283: * Save a copy of the IP header in case we want restore it
284: * for sending an ICMP error message in response.
285: */
286: ip = mtod(m, struct ip *);
287: save_ip = *ip;
288: save_ip.ip_len += iphlen;
289:
290: /*
291: * Checksum extended TCP header and data.
292: */
293: tlen = ((struct ip *)ti)->ip_len;
294: tcpiphdr2qlink(ti)->next = tcpiphdr2qlink(ti)->prev = 0;
295: memset(&ti->ti_i.ih_mbuf, 0 , sizeof(struct mbuf_ptr));
296: ti->ti_x1 = 0;
297: ti->ti_len = htons((u_int16_t)tlen);
298: len = sizeof(struct ip) + tlen;
299: /* keep checksum for ICMP reply
300: * ti->ti_sum = cksum(m, len);
301: * if (ti->ti_sum) { */
302: if (cksum(m, len)) {
303: tcpstat.tcps_rcvbadsum++;
304: goto drop;
305: }
306:
307: /*
308: * Check that TCP offset makes sense,
309: * pull out TCP options and adjust length. XXX
310: */
311: off = ti->ti_off << 2;
312: if (off < sizeof(struct tcphdr) || off > tlen) {
313: tcpstat.tcps_rcvbadoff++;
314: goto drop;
315: }
316: tlen -= off;
317: ti->ti_len = tlen;
318: if (off > sizeof(struct tcphdr)) {
319: optlen = off - sizeof(struct tcphdr);
320: optp = mtod(m, caddr_t) + sizeof(struct tcpiphdr);
321:
322: /*
323: * Do quick retrieval of timestamp options ("options
324: * prediction?"). If timestamp is the only option and it's
325: * formatted as recommended in RFC 1323 appendix A, we
326: * quickly get the values now and not bother calling
327: * tcp_dooptions(), etc.
328: */
329: /* if ((optlen == TCPOLEN_TSTAMP_APPA ||
330: * (optlen > TCPOLEN_TSTAMP_APPA &&
331: * optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
332: * *(u_int32_t *)optp == htonl(TCPOPT_TSTAMP_HDR) &&
333: * (ti->ti_flags & TH_SYN) == 0) {
334: * ts_present = 1;
335: * ts_val = ntohl(*(u_int32_t *)(optp + 4));
336: * ts_ecr = ntohl(*(u_int32_t *)(optp + 8));
337: * optp = NULL; / * we've parsed the options * /
338: * }
339: */
340: }
341: tiflags = ti->ti_flags;
342:
343: /*
344: * Convert TCP protocol specific fields to host format.
345: */
346: NTOHL(ti->ti_seq);
347: NTOHL(ti->ti_ack);
348: NTOHS(ti->ti_win);
349: NTOHS(ti->ti_urp);
350:
351: /*
352: * Drop TCP, IP headers and TCP options.
353: */
354: m->m_data += sizeof(struct tcpiphdr) + off - sizeof(struct tcphdr);
355: m->m_len -= sizeof(struct tcpiphdr) + off - sizeof(struct tcphdr);
356:
357: /*
358: * Locate pcb for segment.
359: */
360: findso:
361: so = tcp_last_so;
362: if (so->so_fport != ti->ti_dport ||
363: so->so_lport != ti->ti_sport ||
364: so->so_laddr.s_addr != ti->ti_src.s_addr ||
365: so->so_faddr.s_addr != ti->ti_dst.s_addr) {
366: so = solookup(&tcb, ti->ti_src, ti->ti_sport,
367: ti->ti_dst, ti->ti_dport);
368: if (so)
369: tcp_last_so = so;
370: ++tcpstat.tcps_socachemiss;
371: }
372:
373: /*
374: * If the state is CLOSED (i.e., TCB does not exist) then
375: * all data in the incoming segment is discarded.
376: * If the TCB exists but is in CLOSED state, it is embryonic,
377: * but should either do a listen or a connect soon.
378: *
379: * state == CLOSED means we've done socreate() but haven't
380: * attached it to a protocol yet...
381: *
382: * XXX If a TCB does not exist, and the TH_SYN flag is
383: * the only flag set, then create a session, mark it
384: * as if it was LISTENING, and continue...
385: */
386: if (so == 0) {
387: if ((tiflags & (TH_SYN | TH_FIN | TH_RST | TH_URG | TH_ACK)) != TH_SYN)
388: goto dropwithreset;
389:
390: if ((so = socreate()) == NULL)
391: goto dropwithreset;
392: if (tcp_attach(so) < 0) {
393: free(so); /* Not sofree (if it failed, it's not insqued) */
394: goto dropwithreset;
395: }
396:
397: sbreserve(&so->so_snd, tcp_sndspace);
398: sbreserve(&so->so_rcv, tcp_rcvspace);
399:
400: /* tcp_last_so = so; */ /* XXX ? */
401: /* tp = sototcpcb(so); */
402:
403: so->so_laddr = ti->ti_src;
404: so->so_lport = ti->ti_sport;
405: so->so_faddr = ti->ti_dst;
406: so->so_fport = ti->ti_dport;
407:
408: if ((so->so_iptos = tcp_tos(so)) == 0)
409: so->so_iptos = ((struct ip *)ti)->ip_tos;
410:
411: tp = sototcpcb(so);
412: tp->t_state = TCPS_LISTEN;
413: }
414:
415: /*
416: * If this is a still-connecting socket, this probably
417: * a retransmit of the SYN. Whether it's a retransmit SYN
418: * or something else, we nuke it.
419: */
420: if (so->so_state & SS_ISFCONNECTING)
421: goto drop;
422:
423: tp = sototcpcb(so);
424:
425: /* XXX Should never fail */
426: if (tp == 0)
427: goto dropwithreset;
428: if (tp->t_state == TCPS_CLOSED)
429: goto drop;
430:
431: /* Unscale the window into a 32-bit value. */
432: /* if ((tiflags & TH_SYN) == 0)
433: * tiwin = ti->ti_win << tp->snd_scale;
434: * else
435: */
436: tiwin = ti->ti_win;
437:
438: /*
439: * Segment received on connection.
440: * Reset idle time and keep-alive timer.
441: */
442: tp->t_idle = 0;
443: if (so_options)
444: tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
445: else
446: tp->t_timer[TCPT_KEEP] = tcp_keepidle;
447:
448: /*
449: * Process options if not in LISTEN state,
450: * else do it below (after getting remote address).
451: */
452: if (optp && tp->t_state != TCPS_LISTEN)
453: tcp_dooptions(tp, (u_char *)optp, optlen, ti);
454: /* , */
455: /* &ts_present, &ts_val, &ts_ecr); */
456:
457: /*
458: * Header prediction: check for the two common cases
459: * of a uni-directional data xfer. If the packet has
460: * no control flags, is in-sequence, the window didn't
461: * change and we're not retransmitting, it's a
462: * candidate. If the length is zero and the ack moved
463: * forward, we're the sender side of the xfer. Just
464: * free the data acked & wake any higher level process
465: * that was blocked waiting for space. If the length
466: * is non-zero and the ack didn't move, we're the
467: * receiver side. If we're getting packets in-order
468: * (the reassembly queue is empty), add the data to
469: * the socket buffer and note that we need a delayed ack.
470: *
471: * XXX Some of these tests are not needed
472: * eg: the tiwin == tp->snd_wnd prevents many more
473: * predictions.. with no *real* advantage..
474: */
475: if (tp->t_state == TCPS_ESTABLISHED &&
476: (tiflags & (TH_SYN | TH_FIN | TH_RST | TH_URG | TH_ACK)) == TH_ACK &&
477: /* (!ts_present || TSTMP_GEQ(ts_val, tp->ts_recent)) && */
478: ti->ti_seq == tp->rcv_nxt &&
479: tiwin && tiwin == tp->snd_wnd &&
480: tp->snd_nxt == tp->snd_max) {
481: /*
482: * If last ACK falls within this segment's sequence numbers,
483: * record the timestamp.
484: */
485: /* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
486: * SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
487: * tp->ts_recent_age = tcp_now;
488: * tp->ts_recent = ts_val;
489: * }
490: */
491: if (ti->ti_len == 0) {
492: if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
493: SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
494: tp->snd_cwnd >= tp->snd_wnd) {
495: /*
496: * this is a pure ack for outstanding data.
497: */
498: ++tcpstat.tcps_predack;
499: /* if (ts_present)
500: * tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
501: * else
502: */ if (tp->t_rtt &&
503: SEQ_GT(ti->ti_ack, tp->t_rtseq))
504: tcp_xmit_timer(tp, tp->t_rtt);
505: acked = ti->ti_ack - tp->snd_una;
506: tcpstat.tcps_rcvackpack++;
507: tcpstat.tcps_rcvackbyte += acked;
508: sbdrop(&so->so_snd, acked);
509: tp->snd_una = ti->ti_ack;
510: m_freem(m);
511:
512: /*
513: * If all outstanding data are acked, stop
514: * retransmit timer, otherwise restart timer
515: * using current (possibly backed-off) value.
516: * If process is waiting for space,
517: * wakeup/selwakeup/signal. If data
518: * are ready to send, let tcp_output
519: * decide between more output or persist.
520: */
521: if (tp->snd_una == tp->snd_max)
522: tp->t_timer[TCPT_REXMT] = 0;
523: else if (tp->t_timer[TCPT_PERSIST] == 0)
524: tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
525:
526: /*
527: * There's room in so_snd, sowwakup will read()
528: * from the socket if we can
529: */
530: /* if (so->so_snd.sb_flags & SB_NOTIFY)
531: * sowwakeup(so);
532: */
533: /*
534: * This is called because sowwakeup might have
535: * put data into so_snd. Since we don't so sowwakeup,
536: * we don't need this.. XXX???
537: */
538: if (so->so_snd.sb_cc)
539: (void) tcp_output(tp);
540:
541: return;
542: }
543: }
544: else if (ti->ti_ack == tp->snd_una &&
545: tcpfrag_list_empty(tp) &&
546: ti->ti_len <= sbspace(&so->so_rcv)) {
547: /*
548: * this is a pure, in-sequence data packet
549: * with nothing on the reassembly queue and
550: * we have enough buffer space to take it.
551: */
552: ++tcpstat.tcps_preddat;
553: tp->rcv_nxt += ti->ti_len;
554: tcpstat.tcps_rcvpack++;
555: tcpstat.tcps_rcvbyte += ti->ti_len;
556: /*
557: * Add data to socket buffer.
558: */
559: if (so->so_emu) {
560: if (tcp_emu(so, m)) sbappend(so, m);
561: }
562: else
563: sbappend(so, m);
564:
565: /*
566: * XXX This is called when data arrives. Later, check
567: * if we can actually write() to the socket
568: * XXX Need to check? It's be NON_BLOCKING
569: */
570: /* sorwakeup(so); */
571:
572: /*
573: * If this is a short packet, then ACK now - with Nagel
574: * congestion avoidance sender won't send more until
575: * he gets an ACK.
576: *
577: * It is better to not delay acks at all to maximize
578: * TCP throughput. See RFC 2581.
579: */
580: tp->t_flags |= TF_ACKNOW;
581: tcp_output(tp);
582: return;
583: }
584: } /* header prediction */
585: /*
586: * Calculate amount of space in receive window,
587: * and then do TCP input processing.
588: * Receive window is amount of space in rcv queue,
589: * but not less than advertised window.
590: */
591: { int win;
592: win = sbspace(&so->so_rcv);
593: if (win < 0)
594: win = 0;
595: tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
596: }
597:
598: switch (tp->t_state) {
599:
600: /*
601: * If the state is LISTEN then ignore segment if it contains an RST.
602: * If the segment contains an ACK then it is bad and send a RST.
603: * If it does not contain a SYN then it is not interesting; drop it.
604: * Don't bother responding if the destination was a broadcast.
605: * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
606: * tp->iss, and send a segment:
607: * <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
608: * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
609: * Fill in remote peer address fields if not previously specified.
610: * Enter SYN_RECEIVED state, and process any other fields of this
611: * segment in this state.
612: */
613: case TCPS_LISTEN: {
614:
615: if (tiflags & TH_RST)
616: goto drop;
617: if (tiflags & TH_ACK)
618: goto dropwithreset;
619: if ((tiflags & TH_SYN) == 0)
620: goto drop;
621:
622: /*
623: * This has way too many gotos...
624: * But a bit of spaghetti code never hurt anybody :)
625: */
626:
627: /*
628: * If this is destined for the control address, then flag to
629: * tcp_ctl once connected, otherwise connect
630: */
631: if ((so->so_faddr.s_addr&htonl(0xffffff00)) == special_addr.s_addr) {
632: int lastbyte = ntohl(so->so_faddr.s_addr) & 0xff;
633: if (lastbyte != CTL_ALIAS && lastbyte != CTL_DNS) {
634: #if 0
635: if (lastbyte == CTL_CMD || lastbyte == CTL_EXEC) {
636: /* Command or exec adress */
637: so->so_state |= SS_CTL;
638: }
639: else
640: #endif
641: {
642: /* May be an add exec */
643: struct ex_list *ex_ptr;
644: for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
645: if (ex_ptr->ex_fport == so->so_fport &&
646: lastbyte == ex_ptr->ex_addr) {
647: so->so_state |= SS_CTL;
648: break;
649: }
650: }
651: }
652: if (so->so_state & SS_CTL) goto cont_input;
653: }
654: /* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */
655: }
656:
657: if (so->so_emu & EMU_NOCONNECT) {
658: so->so_emu &= ~EMU_NOCONNECT;
659: goto cont_input;
660: }
661:
662: if (tcp_fconnect(so) == -1) {
663: int error = WSAGetLastError();
664: if ((error != WSAEINPROGRESS) && (error != WSAEWOULDBLOCK)) {
665: u_char code = ICMP_UNREACH_NET;
666: DEBUG_MISC((dfd, " tcp fconnect errno = %d-%s\n",
667: errno, strerror(errno)));
668: if (error == WSAECONNREFUSED) {
669: /* ACK the SYN, send RST to refuse the connection */
670: tcp_respond(tp, ti, m, ti->ti_seq + 1, (tcp_seq)0,
671: TH_RST | TH_ACK);
672: }
673: else {
674: if (error == WSAEHOSTUNREACH) code = ICMP_UNREACH_HOST;
675: HTONL(ti->ti_seq); /* restore tcp header */
676: HTONL(ti->ti_ack);
677: HTONS(ti->ti_win);
678: HTONS(ti->ti_urp);
679: m->m_data -= sizeof(struct tcpiphdr) + off - sizeof(struct tcphdr);
680: m->m_len += sizeof(struct tcpiphdr) + off - sizeof(struct tcphdr);
681: *ip = save_ip;
682: icmp_error(m, ICMP_UNREACH, code, 0, strerror(errno));
683: }
684: tp = tcp_close(tp);
685: m_free(m);
686: return;
687: }
688: }
689:
690: /*
691: * Haven't connected yet, save the current mbuf
692: * and ti, and return
693: * XXX Some OS's don't tell us whether the connect()
694: * succeeded or not. So we must time it out.
695: */
696: so->so_m = m;
697: so->so_ti = ti;
698: tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
699: tp->t_state = TCPS_SYN_RECEIVED;
700: return;
701:
702: cont_conn:
703: /* m==NULL
704: * Check if the connect succeeded
705: */
706: if (so->so_state & SS_NOFDREF) {
707: tp = tcp_close(tp);
708: goto dropwithreset;
709: }
710: cont_input:
711: tcp_template(tp);
712:
713: if (optp)
714: tcp_dooptions(tp, (u_char *)optp, optlen, ti);
715: /* , */
716: /* &ts_present, &ts_val, &ts_ecr); */
717:
718: if (iss)
719: tp->iss = iss;
720: else
721: tp->iss = tcp_iss;
722: tcp_iss += TCP_ISSINCR / 2;
723: tp->irs = ti->ti_seq;
724: tcp_sendseqinit(tp);
725: tcp_rcvseqinit(tp);
726: tp->t_flags |= TF_ACKNOW;
727: tp->t_state = TCPS_SYN_RECEIVED;
728: tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
729: tcpstat.tcps_accepts++;
730: goto trimthenstep6;
731: } /* case TCPS_LISTEN */
732:
733: /*
734: * If the state is SYN_SENT:
735: * if seg contains an ACK, but not for our SYN, drop the input.
736: * if seg contains a RST, then drop the connection.
737: * if seg does not contain SYN, then drop it.
738: * Otherwise this is an acceptable SYN segment
739: * initialize tp->rcv_nxt and tp->irs
740: * if seg contains ack then advance tp->snd_una
741: * if SYN has been acked change to ESTABLISHED else SYN_RCVD state
742: * arrange for segment to be acked (eventually)
743: * continue processing rest of data/controls, beginning with URG
744: */
745: case TCPS_SYN_SENT:
746: if ((tiflags & TH_ACK) &&
747: (SEQ_LEQ(ti->ti_ack, tp->iss) ||
748: SEQ_GT(ti->ti_ack, tp->snd_max)))
749: goto dropwithreset;
750:
751: if (tiflags & TH_RST) {
752: if (tiflags & TH_ACK)
753: tp = tcp_drop(tp, 0); /* XXX Check t_softerror! */
754: goto drop;
755: }
756:
757: if ((tiflags & TH_SYN) == 0)
758: goto drop;
759: if (tiflags & TH_ACK) {
760: tp->snd_una = ti->ti_ack;
761: if (SEQ_LT(tp->snd_nxt, tp->snd_una))
762: tp->snd_nxt = tp->snd_una;
763: }
764:
765: tp->t_timer[TCPT_REXMT] = 0;
766: tp->irs = ti->ti_seq;
767: tcp_rcvseqinit(tp);
768: tp->t_flags |= TF_ACKNOW;
769: if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
770: tcpstat.tcps_connects++;
771: soisfconnected(so);
772: tp->t_state = TCPS_ESTABLISHED;
773:
774: /* Do window scaling on this connection? */
775: /* if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
776: * (TF_RCVD_SCALE|TF_REQ_SCALE)) {
777: * tp->snd_scale = tp->requested_s_scale;
778: * tp->rcv_scale = tp->request_r_scale;
779: * }
780: */
781: (void)tcp_reass(tp, (struct tcpiphdr *)0,
782: (struct mbuf *)0);
783: /*
784: * if we didn't have to retransmit the SYN,
785: * use its rtt as our initial srtt & rtt var.
786: */
787: if (tp->t_rtt)
788: tcp_xmit_timer(tp, tp->t_rtt);
789: }
790: else
791: tp->t_state = TCPS_SYN_RECEIVED;
792:
793: trimthenstep6:
794: /*
795: * Advance ti->ti_seq to correspond to first data byte.
796: * If data, trim to stay within window,
797: * dropping FIN if necessary.
798: */
799: ti->ti_seq++;
800: if (ti->ti_len > tp->rcv_wnd) {
801: todrop = ti->ti_len - tp->rcv_wnd;
802: m_adj(m, -todrop);
803: ti->ti_len = tp->rcv_wnd;
804: tiflags &= ~TH_FIN;
805: tcpstat.tcps_rcvpackafterwin++;
806: tcpstat.tcps_rcvbyteafterwin += todrop;
807: }
808: tp->snd_wl1 = ti->ti_seq - 1;
809: tp->rcv_up = ti->ti_seq;
810: goto step6;
811: } /* switch tp->t_state */
812: /*
813: * States other than LISTEN or SYN_SENT.
814: * First check timestamp, if present.
815: * Then check that at least some bytes of segment are within
816: * receive window. If segment begins before rcv_nxt,
817: * drop leading data (and SYN); if nothing left, just ack.
818: *
819: * RFC 1323 PAWS: If we have a timestamp reply on this segment
820: * and it's less than ts_recent, drop it.
821: */
822: /* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
823: * TSTMP_LT(ts_val, tp->ts_recent)) {
824: *
825: */ /* Check to see if ts_recent is over 24 days old. */
826: /* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
827: */ /*
828: * * Invalidate ts_recent. If this segment updates
829: * * ts_recent, the age will be reset later and ts_recent
830: * * will get a valid value. If it does not, setting
831: * * ts_recent to zero will at least satisfy the
832: * * requirement that zero be placed in the timestamp
833: * * echo reply when ts_recent isn't valid. The
834: * * age isn't reset until we get a valid ts_recent
835: * * because we don't want out-of-order segments to be
836: * * dropped when ts_recent is old.
837: * */
838: /* tp->ts_recent = 0;
839: * } else {
840: * tcpstat.tcps_rcvduppack++;
841: * tcpstat.tcps_rcvdupbyte += ti->ti_len;
842: * tcpstat.tcps_pawsdrop++;
843: * goto dropafterack;
844: * }
845: * }
846: */
847:
848: todrop = tp->rcv_nxt - ti->ti_seq;
849: if (todrop > 0) {
850: if (tiflags & TH_SYN) {
851: tiflags &= ~TH_SYN;
852: ti->ti_seq++;
853: if (ti->ti_urp > 1)
854: ti->ti_urp--;
855: else
856: tiflags &= ~TH_URG;
857: todrop--;
858: }
859: /*
860: * Following if statement from Stevens, vol. 2, p. 960.
861: */
862: if (todrop > ti->ti_len
863: || (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
864: /*
865: * Any valid FIN must be to the left of the window.
866: * At this point the FIN must be a duplicate or out
867: * of sequence; drop it.
868: */
869: tiflags &= ~TH_FIN;
870:
871: /*
872: * Send an ACK to resynchronize and drop any data.
873: * But keep on processing for RST or ACK.
874: */
875: tp->t_flags |= TF_ACKNOW;
876: todrop = ti->ti_len;
877: tcpstat.tcps_rcvduppack++;
878: tcpstat.tcps_rcvdupbyte += todrop;
879: }
880: else {
881: tcpstat.tcps_rcvpartduppack++;
882: tcpstat.tcps_rcvpartdupbyte += todrop;
883: }
884: m_adj(m, todrop);
885: ti->ti_seq += todrop;
886: ti->ti_len -= todrop;
887: if (ti->ti_urp > todrop)
888: ti->ti_urp -= todrop;
889: else {
890: tiflags &= ~TH_URG;
891: ti->ti_urp = 0;
892: }
893: }
894: /*
895: * If new data are received on a connection after the
896: * user processes are gone, then RST the other end.
897: */
898: if ((so->so_state & SS_NOFDREF) &&
899: tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
900: tp = tcp_close(tp);
901: tcpstat.tcps_rcvafterclose++;
902: goto dropwithreset;
903: }
904:
905: /*
906: * If segment ends after window, drop trailing data
907: * (and PUSH and FIN); if nothing left, just ACK.
908: */
909: todrop = (ti->ti_seq + ti->ti_len) - (tp->rcv_nxt + tp->rcv_wnd);
910: if (todrop > 0) {
911: tcpstat.tcps_rcvpackafterwin++;
912: if (todrop >= ti->ti_len) {
913: tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
914: /*
915: * If a new connection request is received
916: * while in TIME_WAIT, drop the old connection
917: * and start over if the sequence numbers
918: * are above the previous ones.
919: */
920: if (tiflags & TH_SYN &&
921: tp->t_state == TCPS_TIME_WAIT &&
922: SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
923: iss = tp->rcv_nxt + TCP_ISSINCR;
924: tp = tcp_close(tp);
925: goto findso;
926: }
927: /*
928: * If window is closed can only take segments at
929: * window edge, and have to drop data and PUSH from
930: * incoming segments. Continue processing, but
931: * remember to ack. Otherwise, drop segment
932: * and ack.
933: */
934: if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
935: tp->t_flags |= TF_ACKNOW;
936: tcpstat.tcps_rcvwinprobe++;
937: }
938: else
939: goto dropafterack;
940: }
941: else
942: tcpstat.tcps_rcvbyteafterwin += todrop;
943: m_adj(m, -todrop);
944: ti->ti_len -= todrop;
945: tiflags &= ~(TH_PUSH | TH_FIN);
946: }
947:
948: /*
949: * If last ACK falls within this segment's sequence numbers,
950: * record its timestamp.
951: */
952: /* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
953: * SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
954: * ((tiflags & (TH_SYN|TH_FIN)) != 0))) {
955: * tp->ts_recent_age = tcp_now;
956: * tp->ts_recent = ts_val;
957: * }
958: */
959:
960: /*
961: * If the RST bit is set examine the state:
962: * SYN_RECEIVED STATE:
963: * If passive open, return to LISTEN state.
964: * If active open, inform user that connection was refused.
965: * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
966: * Inform user that connection was reset, and close tcb.
967: * CLOSING, LAST_ACK, TIME_WAIT STATES
968: * Close the tcb.
969: */
970: if (tiflags&TH_RST) switch (tp->t_state) {
971:
972: case TCPS_SYN_RECEIVED:
973: /* so->so_error = ECONNREFUSED; */
974: goto close;
975:
976: case TCPS_ESTABLISHED:
977: case TCPS_FIN_WAIT_1:
978: case TCPS_FIN_WAIT_2:
979: case TCPS_CLOSE_WAIT:
980: /* so->so_error = ECONNRESET; */
981: close:
982: tp->t_state = TCPS_CLOSED;
983: tcpstat.tcps_drops++;
984: tp = tcp_close(tp);
985: goto drop;
986:
987: case TCPS_CLOSING:
988: case TCPS_LAST_ACK:
989: case TCPS_TIME_WAIT:
990: tp = tcp_close(tp);
991: goto drop;
992: }
993:
994: /*
995: * If a SYN is in the window, then this is an
996: * error and we send an RST and drop the connection.
997: */
998: if (tiflags & TH_SYN) {
999: tp = tcp_drop(tp, 0);
1000: goto dropwithreset;
1001: }
1002:
1003: /*
1004: * If the ACK bit is off we drop the segment and return.
1005: */
1006: if ((tiflags & TH_ACK) == 0) goto drop;
1007:
1008: /*
1009: * Ack processing.
1010: */
1011: switch (tp->t_state) {
1012: /*
1013: * In SYN_RECEIVED state if the ack ACKs our SYN then enter
1014: * ESTABLISHED state and continue processing, otherwise
1015: * send an RST. una<=ack<=max
1016: */
1017: case TCPS_SYN_RECEIVED:
1018:
1019: if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
1020: SEQ_GT(ti->ti_ack, tp->snd_max))
1021: goto dropwithreset;
1022: tcpstat.tcps_connects++;
1023: tp->t_state = TCPS_ESTABLISHED;
1024: /*
1025: * The sent SYN is ack'ed with our sequence number +1
1026: * The first data byte already in the buffer will get
1027: * lost if no correction is made. This is only needed for
1028: * SS_CTL since the buffer is empty otherwise.
1029: * tp->snd_una++; or:
1030: */
1031: tp->snd_una = ti->ti_ack;
1032: if (so->so_state & SS_CTL) {
1033: /* So tcp_ctl reports the right state */
1034: ret = tcp_ctl(so);
1035: if (ret == 1) {
1036: soisfconnected(so);
1037: so->so_state &= ~SS_CTL; /* success XXX */
1038: }
1039: else if (ret == 2) {
1040: so->so_state = SS_NOFDREF; /* CTL_CMD */
1041: }
1042: else {
1043: needoutput = 1;
1044: tp->t_state = TCPS_FIN_WAIT_1;
1045: }
1046: }
1047: else {
1048: soisfconnected(so);
1049: }
1050:
1051: /* Do window scaling? */
1052: /* if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
1053: * (TF_RCVD_SCALE|TF_REQ_SCALE)) {
1054: * tp->snd_scale = tp->requested_s_scale;
1055: * tp->rcv_scale = tp->request_r_scale;
1056: * }
1057: */
1058: (void)tcp_reass(tp, (struct tcpiphdr *)0, (struct mbuf *)0);
1059: tp->snd_wl1 = ti->ti_seq - 1;
1060: /* Avoid ack processing; snd_una==ti_ack => dup ack */
1061: goto synrx_to_est;
1062: /* fall into ... */
1063:
1064: /*
1065: * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1066: * ACKs. If the ack is in the range
1067: * tp->snd_una < ti->ti_ack <= tp->snd_max
1068: * then advance tp->snd_una to ti->ti_ack and drop
1069: * data from the retransmission queue. If this ACK reflects
1070: * more up to date window information we update our window information.
1071: */
1072: case TCPS_ESTABLISHED:
1073: case TCPS_FIN_WAIT_1:
1074: case TCPS_FIN_WAIT_2:
1075: case TCPS_CLOSE_WAIT:
1076: case TCPS_CLOSING:
1077: case TCPS_LAST_ACK:
1078: case TCPS_TIME_WAIT:
1079:
1080: if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
1081: if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
1082: tcpstat.tcps_rcvdupack++;
1083: DEBUG_MISC((dfd, " dup ack m = %lx so = %lx \n",
1084: (long)m, (long)so));
1085: /*
1086: * If we have outstanding data (other than
1087: * a window probe), this is a completely
1088: * duplicate ack (ie, window info didn't
1089: * change), the ack is the biggest we've
1090: * seen and we've seen exactly our rexmt
1091: * threshold of them, assume a packet
1092: * has been dropped and retransmit it.
1093: * Kludge snd_nxt & the congestion
1094: * window so we send only this one
1095: * packet.
1096: *
1097: * We know we're losing at the current
1098: * window size so do congestion avoidance
1099: * (set ssthresh to half the current window
1100: * and pull our congestion window back to
1101: * the new ssthresh).
1102: *
1103: * Dup acks mean that packets have left the
1104: * network (they're now cached at the receiver)
1105: * so bump cwnd by the amount in the receiver
1106: * to keep a constant cwnd packets in the
1107: * network.
1108: */
1109: if (tp->t_timer[TCPT_REXMT] == 0 ||
1110: ti->ti_ack != tp->snd_una)
1111: tp->t_dupacks = 0;
1112: else if (++tp->t_dupacks == tcprexmtthresh) {
1113: tcp_seq onxt = tp->snd_nxt;
1114: u_int win =
1115: min(tp->snd_wnd, tp->snd_cwnd) / 2 /
1116: tp->t_maxseg;
1117:
1118: if (win < 2)
1119: win = 2;
1120: tp->snd_ssthresh = win * tp->t_maxseg;
1121: tp->t_timer[TCPT_REXMT] = 0;
1122: tp->t_rtt = 0;
1123: tp->snd_nxt = ti->ti_ack;
1124: tp->snd_cwnd = tp->t_maxseg;
1125: (void)tcp_output(tp);
1126: tp->snd_cwnd = tp->snd_ssthresh +
1127: tp->t_maxseg * tp->t_dupacks;
1128: if (SEQ_GT(onxt, tp->snd_nxt))
1129: tp->snd_nxt = onxt;
1130: goto drop;
1131: }
1132: else if (tp->t_dupacks > tcprexmtthresh) {
1133: tp->snd_cwnd += tp->t_maxseg;
1134: (void)tcp_output(tp);
1135: goto drop;
1136: }
1137: }
1138: else
1139: tp->t_dupacks = 0;
1140: break;
1141: }
1142: synrx_to_est:
1143: /*
1144: * If the congestion window was inflated to account
1145: * for the other side's cached packets, retract it.
1146: */
1147: if (tp->t_dupacks > tcprexmtthresh &&
1148: tp->snd_cwnd > tp->snd_ssthresh)
1149: tp->snd_cwnd = tp->snd_ssthresh;
1150: tp->t_dupacks = 0;
1151: if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
1152: tcpstat.tcps_rcvacktoomuch++;
1153: goto dropafterack;
1154: }
1155: acked = ti->ti_ack - tp->snd_una;
1156: tcpstat.tcps_rcvackpack++;
1157: tcpstat.tcps_rcvackbyte += acked;
1158:
1159: /*
1160: * If we have a timestamp reply, update smoothed
1161: * round trip time. If no timestamp is present but
1162: * transmit timer is running and timed sequence
1163: * number was acked, update smoothed round trip time.
1164: * Since we now have an rtt measurement, cancel the
1165: * timer backoff (cf., Phil Karn's retransmit alg.).
1166: * Recompute the initial retransmit timer.
1167: */
1168: /* if (ts_present)
1169: * tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1170: * else
1171: */
1172: if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1173: tcp_xmit_timer(tp, tp->t_rtt);
1174:
1175: /*
1176: * If all outstanding data is acked, stop retransmit
1177: * timer and remember to restart (more output or persist).
1178: * If there is more data to be acked, restart retransmit
1179: * timer, using current (possibly backed-off) value.
1180: */
1181: if (ti->ti_ack == tp->snd_max) {
1182: tp->t_timer[TCPT_REXMT] = 0;
1183: needoutput = 1;
1184: }
1185: else if (tp->t_timer[TCPT_PERSIST] == 0)
1186: tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1187: /*
1188: * When new data is acked, open the congestion window.
1189: * If the window gives us less than ssthresh packets
1190: * in flight, open exponentially (maxseg per packet).
1191: * Otherwise open linearly: maxseg per window
1192: * (maxseg^2 / cwnd per packet).
1193: */
1194: {
1195: register u_int cw = tp->snd_cwnd;
1196: register u_int incr = tp->t_maxseg;
1197:
1198: if (cw > tp->snd_ssthresh)
1199: incr = incr * incr / cw;
1200: tp->snd_cwnd = min(cw + incr, (u_int32_t) (TCP_MAXWIN << tp->snd_scale));
1201: }
1202: if (acked > so->so_snd.sb_cc) {
1203: tp->snd_wnd -= so->so_snd.sb_cc;
1204: sbdrop(&so->so_snd, so->so_snd.sb_cc);
1205: ourfinisacked = 1;
1206: }
1207: else {
1208: sbdrop(&so->so_snd, acked);
1209: tp->snd_wnd -= acked;
1210: ourfinisacked = 0;
1211: }
1212: /*
1213: * XXX sowwakup is called when data is acked and there's room for
1214: * for more data... it should read() the socket
1215: */
1216: /* if (so->so_snd.sb_flags & SB_NOTIFY)
1217: * sowwakeup(so);
1218: */
1219: tp->snd_una = ti->ti_ack;
1220: if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1221: tp->snd_nxt = tp->snd_una;
1222:
1223: switch (tp->t_state) {
1224:
1225: /*
1226: * In FIN_WAIT_1 STATE in addition to the processing
1227: * for the ESTABLISHED state if our FIN is now acknowledged
1228: * then enter FIN_WAIT_2.
1229: */
1230: case TCPS_FIN_WAIT_1:
1231: if (ourfinisacked) {
1232: /*
1233: * If we can't receive any more
1234: * data, then closing user can proceed.
1235: * Starting the timer is contrary to the
1236: * specification, but if we don't get a FIN
1237: * we'll hang forever.
1238: */
1239: if (so->so_state & SS_FCANTRCVMORE) {
1240: soisfdisconnected(so);
1241: tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1242: }
1243: tp->t_state = TCPS_FIN_WAIT_2;
1244: }
1245: break;
1246:
1247: /*
1248: * In CLOSING STATE in addition to the processing for
1249: * the ESTABLISHED state if the ACK acknowledges our FIN
1250: * then enter the TIME-WAIT state, otherwise ignore
1251: * the segment.
1252: */
1253: case TCPS_CLOSING:
1254: if (ourfinisacked) {
1255: tp->t_state = TCPS_TIME_WAIT;
1256: tcp_canceltimers(tp);
1257: tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1258: soisfdisconnected(so);
1259: }
1260: break;
1261:
1262: /*
1263: * In LAST_ACK, we may still be waiting for data to drain
1264: * and/or to be acked, as well as for the ack of our FIN.
1265: * If our FIN is now acknowledged, delete the TCB,
1266: * enter the closed state and return.
1267: */
1268: case TCPS_LAST_ACK:
1269: if (ourfinisacked) {
1270: tp = tcp_close(tp);
1271: goto drop;
1272: }
1273: break;
1274:
1275: /*
1276: * In TIME_WAIT state the only thing that should arrive
1277: * is a retransmission of the remote FIN. Acknowledge
1278: * it and restart the finack timer.
1279: */
1280: case TCPS_TIME_WAIT:
1281: tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1282: goto dropafterack;
1283: }
1284: } /* switch(tp->t_state) */
1285:
1286: step6:
1287: /*
1288: * Update window information.
1289: * Don't look at window if no ACK: TAC's send garbage on first SYN.
1290: */
1291: if ((tiflags & TH_ACK) &&
1292: (SEQ_LT(tp->snd_wl1, ti->ti_seq) ||
1293: (tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) ||
1294: (tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
1295: /* keep track of pure window updates */
1296: if (ti->ti_len == 0 &&
1297: tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd)
1298: tcpstat.tcps_rcvwinupd++;
1299: tp->snd_wnd = tiwin;
1300: tp->snd_wl1 = ti->ti_seq;
1301: tp->snd_wl2 = ti->ti_ack;
1302: if (tp->snd_wnd > tp->max_sndwnd)
1303: tp->max_sndwnd = tp->snd_wnd;
1304: needoutput = 1;
1305: }
1306:
1307: /*
1308: * Process segments with URG.
1309: */
1310: if ((tiflags & TH_URG) && ti->ti_urp &&
1311: TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1312: /*
1313: * This is a kludge, but if we receive and accept
1314: * random urgent pointers, we'll crash in
1315: * soreceive. It's hard to imagine someone
1316: * actually wanting to send this much urgent data.
1317: */
1318: if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
1319: ti->ti_urp = 0;
1320: tiflags &= ~TH_URG;
1321: goto dodata;
1322: }
1323: /*
1324: * If this segment advances the known urgent pointer,
1325: * then mark the data stream. This should not happen
1326: * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1327: * a FIN has been received from the remote side.
1328: * In these states we ignore the URG.
1329: *
1330: * According to RFC961 (Assigned Protocols),
1331: * the urgent pointer points to the last octet
1332: * of urgent data. We continue, however,
1333: * to consider it to indicate the first octet
1334: * of data past the urgent section as the original
1335: * spec states (in one of two places).
1336: */
1337: if (SEQ_GT(ti->ti_seq + ti->ti_urp, tp->rcv_up)) {
1338: tp->rcv_up = ti->ti_seq + ti->ti_urp;
1339: so->so_urgc = so->so_rcv.sb_cc +
1340: (tp->rcv_up - tp->rcv_nxt); /* -1; */
1341: tp->rcv_up = ti->ti_seq + ti->ti_urp;
1342:
1343: }
1344: }
1345: else
1346: /*
1347: * If no out of band data is expected,
1348: * pull receive urgent pointer along
1349: * with the receive window.
1350: */
1351: if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1352: tp->rcv_up = tp->rcv_nxt;
1353: dodata:
1354:
1355: /*
1356: * Process the segment text, merging it into the TCP sequencing queue,
1357: * and arranging for acknowledgment of receipt if necessary.
1358: * This process logically involves adjusting tp->rcv_wnd as data
1359: * is presented to the user (this happens in tcp_usrreq.c,
1360: * case PRU_RCVD). If a FIN has already been received on this
1361: * connection then we just ignore the text.
1362: */
1363: if ((ti->ti_len || (tiflags&TH_FIN)) &&
1364: TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1365: TCP_REASS(tp, ti, m, so, tiflags);
1366: /*
1367: * Note the amount of data that peer has sent into
1368: * our window, in order to estimate the sender's
1369: * buffer size.
1370: */
1371: len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt);
1372: }
1373: else {
1374: m_free(m);
1375: tiflags &= ~TH_FIN;
1376: }
1377:
1378: /*
1379: * If FIN is received ACK the FIN and let the user know
1380: * that the connection is closing.
1381: */
1382: if (tiflags & TH_FIN) {
1383: if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1384: /*
1385: * If we receive a FIN we can't send more data,
1386: * set it SS_FDRAIN
1387: * Shutdown the socket if there is no rx data in the
1388: * buffer.
1389: * soread() is called on completion of shutdown() and
1390: * will got to TCPS_LAST_ACK, and use tcp_output()
1391: * to send the FIN.
1392: */
1393: /* sofcantrcvmore(so); */
1394: sofwdrain(so);
1395:
1396: tp->t_flags |= TF_ACKNOW;
1397: tp->rcv_nxt++;
1398: }
1399: switch (tp->t_state) {
1400:
1401: /*
1402: * In SYN_RECEIVED and ESTABLISHED STATES
1403: * enter the CLOSE_WAIT state.
1404: */
1405: case TCPS_SYN_RECEIVED:
1406: case TCPS_ESTABLISHED:
1407: if (so->so_emu == EMU_CTL) /* no shutdown on socket */
1408: tp->t_state = TCPS_LAST_ACK;
1409: else
1410: tp->t_state = TCPS_CLOSE_WAIT;
1411: break;
1412:
1413: /*
1414: * If still in FIN_WAIT_1 STATE FIN has not been acked so
1415: * enter the CLOSING state.
1416: */
1417: case TCPS_FIN_WAIT_1:
1418: tp->t_state = TCPS_CLOSING;
1419: break;
1420:
1421: /*
1422: * In FIN_WAIT_2 state enter the TIME_WAIT state,
1423: * starting the time-wait timer, turning off the other
1424: * standard timers.
1425: */
1426: case TCPS_FIN_WAIT_2:
1427: tp->t_state = TCPS_TIME_WAIT;
1428: tcp_canceltimers(tp);
1429: tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1430: soisfdisconnected(so);
1431: break;
1432:
1433: /*
1434: * In TIME_WAIT state restart the 2 MSL time_wait timer.
1435: */
1436: case TCPS_TIME_WAIT:
1437: tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1438: break;
1439: }
1440: }
1441:
1442: /*
1443: * If this is a small packet, then ACK now - with Nagel
1444: * congestion avoidance sender won't send more until
1445: * he gets an ACK.
1446: *
1447: * See above.
1448: */
1449: /* if (ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg) {
1450: */
1451: /* if ((ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg &&
1452: * (so->so_iptos & IPTOS_LOWDELAY) == 0) ||
1453: * ((so->so_iptos & IPTOS_LOWDELAY) &&
1454: * ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
1455: */
1456: if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
1457: ((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
1458: tp->t_flags |= TF_ACKNOW;
1459: }
1460:
1461: /*
1462: * Return any desired output.
1463: */
1464: if (needoutput || (tp->t_flags & TF_ACKNOW)) {
1465: (void)tcp_output(tp);
1466: }
1467: return;
1468:
1469: dropafterack:
1470: /*
1471: * Generate an ACK dropping incoming segment if it occupies
1472: * sequence space, where the ACK reflects our state.
1473: */
1474: if (tiflags & TH_RST)
1475: goto drop;
1476: m_freem(m);
1477: tp->t_flags |= TF_ACKNOW;
1478: (void)tcp_output(tp);
1479: return;
1480:
1481: dropwithreset:
1482: /* reuses m if m!=NULL, m_free() unnecessary */
1483: if (tiflags & TH_ACK)
1484: tcp_respond(tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1485: else {
1486: if (tiflags & TH_SYN) ti->ti_len++;
1487: tcp_respond(tp, ti, m, ti->ti_seq + ti->ti_len, (tcp_seq)0,
1488: TH_RST | TH_ACK);
1489: }
1490:
1491: return;
1492:
1493: drop:
1494: /*
1495: * Drop space held by incoming segment and return.
1496: */
1497: m_free(m);
1498:
1499: return;
1500: }
1501:
1502: /* , ts_present, ts_val, ts_ecr) */
1503: /* int *ts_present;
1504: * u_int32_t *ts_val, *ts_ecr;
1505: */
1506: void
1507: tcp_dooptions(struct tcpcb *tp, u_char *cp, int cnt, struct tcpiphdr *ti)
1508: {
1509: u_int16_t mss;
1510: int opt, optlen;
1511:
1512: DEBUG_CALL("tcp_dooptions");
1513: DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt));
1514:
1515: for (; cnt > 0; cnt -= optlen, cp += optlen) {
1516: opt = cp[0];
1517: if (opt == TCPOPT_EOL)
1518: break;
1519: if (opt == TCPOPT_NOP)
1520: optlen = 1;
1521: else {
1522: optlen = cp[1];
1523: if (optlen <= 0)
1524: break;
1525: }
1526: switch (opt) {
1527:
1528: default:
1529: continue;
1530:
1531: case TCPOPT_MAXSEG:
1532: if (optlen != TCPOLEN_MAXSEG)
1533: continue;
1534: if (!(ti->ti_flags & TH_SYN))
1535: continue;
1536: memcpy((char *) &mss, (char *) cp + 2, sizeof(mss));
1537: NTOHS(mss);
1538: tcp_mss(tp, mss); /* sets t_maxseg */
1539: break;
1540:
1541: /* case TCPOPT_WINDOW:
1542: * if (optlen != TCPOLEN_WINDOW)
1543: * continue;
1544: * if (!(ti->ti_flags & TH_SYN))
1545: * continue;
1546: * tp->t_flags |= TF_RCVD_SCALE;
1547: * tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1548: * break;
1549: */
1550: /* case TCPOPT_TIMESTAMP:
1551: * if (optlen != TCPOLEN_TIMESTAMP)
1552: * continue;
1553: * *ts_present = 1;
1554: * memcpy((char *) ts_val, (char *)cp + 2, sizeof(*ts_val));
1555: * NTOHL(*ts_val);
1556: * memcpy((char *) ts_ecr, (char *)cp + 6, sizeof(*ts_ecr));
1557: * NTOHL(*ts_ecr);
1558: *
1559: */ /*
1560: * * A timestamp received in a SYN makes
1561: * * it ok to send timestamp requests and replies.
1562: * */
1563: /* if (ti->ti_flags & TH_SYN) {
1564: * tp->t_flags |= TF_RCVD_TSTMP;
1565: * tp->ts_recent = *ts_val;
1566: * tp->ts_recent_age = tcp_now;
1567: * }
1568: */ break;
1569: }
1570: }
1571: }
1572:
1573:
1574: /*
1575: * Pull out of band byte out of a segment so
1576: * it doesn't appear in the user's data queue.
1577: * It is still reflected in the segment length for
1578: * sequencing purposes.
1579: */
1580:
1581: #ifdef notdef
1582:
1583: void tcp_pulloutofband(struct socket *so, struct tcpiphdr *ti, register struct mbuf *m)
1584: {
1585: int cnt = ti->ti_urp - 1;
1586:
1587: while (cnt >= 0) {
1588: if (m->m_len > cnt) {
1589: char *cp = mtod(m, caddr_t) + cnt;
1590: struct tcpcb *tp = sototcpcb(so);
1591:
1592: tp->t_iobc = *cp;
1593: tp->t_oobflags |= TCPOOB_HAVEDATA;
1594: memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1595: m->m_len--;
1596: return;
1597: }
1598: cnt -= m->m_len;
1599: m = m->m_next; /* XXX WRONG! Fix it! */
1600: if (m == 0)
1601: break;
1602: }
1603: panic("tcp_pulloutofband");
1604: }
1605:
1606: #endif /* notdef */
1607:
1608: /*
1609: * Collect new round-trip time estimate
1610: * and update averages and current timeout.
1611: */
1612:
1613: void tcp_xmit_timer(register struct tcpcb *tp, int rtt)
1614: {
1615: register short delta;
1616:
1617: DEBUG_CALL("tcp_xmit_timer");
1618: DEBUG_ARG("tp = %lx", (long)tp);
1619: DEBUG_ARG("rtt = %d", rtt);
1620:
1621: tcpstat.tcps_rttupdated++;
1622: if (tp->t_srtt != 0) {
1623: /*
1624: * srtt is stored as fixed point with 3 bits after the
1625: * binary point (i.e., scaled by 8). The following magic
1626: * is equivalent to the smoothing algorithm in rfc793 with
1627: * an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1628: * point). Adjust rtt to origin 0.
1629: */
1630: delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1631: if ((tp->t_srtt += delta) <= 0)
1632: tp->t_srtt = 1;
1633: /*
1634: * We accumulate a smoothed rtt variance (actually, a
1635: * smoothed mean difference), then set the retransmit
1636: * timer to smoothed rtt + 4 times the smoothed variance.
1637: * rttvar is stored as fixed point with 2 bits after the
1638: * binary point (scaled by 4). The following is
1639: * equivalent to rfc793 smoothing with an alpha of .75
1640: * (rttvar = rttvar*3/4 + |delta| / 4). This replaces
1641: * rfc793's wired-in beta.
1642: */
1643: if (delta < 0)
1644: delta = -delta;
1645: delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1646: if ((tp->t_rttvar += delta) <= 0)
1647: tp->t_rttvar = 1;
1648: } else {
1649: /*
1650: * No rtt measurement yet - use the unsmoothed rtt.
1651: * Set the variance to half the rtt (so our first
1652: * retransmit happens at 3*rtt).
1653: */
1654: tp->t_srtt = rtt << TCP_RTT_SHIFT;
1655: tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1656: }
1657: tp->t_rtt = 0;
1658: tp->t_rxtshift = 0;
1659:
1660: /*
1661: * the retransmit should happen at rtt + 4 * rttvar.
1662: * Because of the way we do the smoothing, srtt and rttvar
1663: * will each average +1/2 tick of bias. When we compute
1664: * the retransmit timer, we want 1/2 tick of rounding and
1665: * 1 extra tick because of +-1/2 tick uncertainty in the
1666: * firing of the timer. The bias will give us exactly the
1667: * 1.5 tick we need. But, because the bias is
1668: * statistical, we have to test that we don't drop below
1669: * the minimum feasible timer (which is 2 ticks).
1670: */
1671: TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1672: (short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1673:
1674: /*
1675: * We received an ack for a packet that wasn't retransmitted;
1676: * it is probably safe to discard any error indications we've
1677: * received recently. This isn't quite right, but close enough
1678: * for now (a route might have failed after we sent a segment,
1679: * and the return path might not be symmetrical).
1680: */
1681: tp->t_softerror = 0;
1682: }
1683:
1684: /*
1685: * Determine a reasonable value for maxseg size.
1686: * If the route is known, check route for mtu.
1687: * If none, use an mss that can be handled on the outgoing
1688: * interface without forcing IP to fragment; if bigger than
1689: * an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1690: * to utilize large mbufs. If no route is found, route has no mtu,
1691: * or the destination isn't local, use a default, hopefully conservative
1692: * size (usually 512 or the default IP max size, but no more than the mtu
1693: * of the interface), as we can't discover anything about intervening
1694: * gateways or networks. We also initialize the congestion/slow start
1695: * window to be a single segment if the destination isn't local.
1696: * While looking at the routing entry, we also initialize other path-dependent
1697: * parameters from pre-set or cached values in the routing entry.
1698: */
1699:
1700: u_int tcp_mss(register struct tcpcb *tp, u_int offer)
1701: {
1702: struct socket *so = tp->t_socket;
1703: u_int mss;
1704:
1705: DEBUG_CALL("tcp_mss");
1706: DEBUG_ARG("tp = %lx", (long)tp);
1707: DEBUG_ARG("offer = %d", offer);
1708:
1709: mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1710: if (offer)
1711: mss = min(mss, offer);
1712: mss = max(mss, 32);
1713: if (mss < tp->t_maxseg || offer != 0)
1714: tp->t_maxseg = mss;
1715:
1716: tp->snd_cwnd = mss;
1717:
1718: sbreserve(&so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1719: sbreserve(&so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1720:
1721: DEBUG_MISC((dfd, " returning mss = %d\n", mss));
1722:
1723: return mss;
1724: }
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.